MICU1 controls spatial membrane potential gradients and guides Ca2+ fluxes within mitochondrial substructures

Mitochondrial ultrastructure represents a pinnacle of form and function, with the inner mitochondrial membrane (IMM) forming isolated pockets of cristae membrane (CM), separated from the inner-boundary membrane (IBM) by cristae junctions (CJ). Applying structured illumination and electron microscopy, a novel and fundamental function of MICU1 in mediating Ca2+ control over spatial membrane potential gradients (SMPGs) between CM and IMS was identified. We unveiled alterations of SMPGs by transient CJ openings when Ca2+ binds to MICU1 resulting in spatial cristae depolarization. This Ca2+/MICU1-mediated plasticity of the CJ further provides the mechanistic bedrock of the biphasic mitochondrial Ca2+ uptake kinetics via the mitochondrial Ca2+ uniporter (MCU) during intracellular Ca2+ release: Initially, high Ca2+ opens CJ via Ca2+/MICU1 and allows instant Ca2+ uptake across the CM through constantly active MCU. Second, MCU disseminates into the IBM, thus establishing Ca2+ uptake across the IBM that circumvents the CM. Under the condition of MICU1 methylation by PRMT1 in aging or cancer, UCP2 that binds to methylated MICU1 destabilizes CJ, disrupts SMPGs, and facilitates fast Ca2+ uptake via the CM.

T he ultrastructure of the inner mitochondrial membrane (IMM) is separated into two parts, the inner boundary membrane (IBM) representing the interface of the IMM with the outer mitochondrial membrane (OMM), and the cristae membrane (CM), forming the membrane invaginations into the mitochondrial matrix 1 . Morphologically, the cristae junctions (CJ), bottleneck membrane structures with a diameter of approximately 17 nm 2,3 , separate IBM and CM from each other. The mitochondrial contact site and cristae organizing system (MICOS) proteins and the scaffold protein optic atrophy 1 (OPA1) stabilize the CJ 4 .
All three IMM compartments CJ, IBM, and CM differ in morphology and composition of characteristic proteins according to their spatio-specific functions. In particular, the IBM hosts the mitochondrial protein uptake complex TIM (translocase of the inner membrane) that interacts with the TOM (translocase of the outer membrane) complex in the OMM to translocate nuclearencoded mitochondrial proteins into the mitochondrial matrix or mitochondrial membranes 5 . The CM contains the four complexes of the respiratory chain 6 . Of those, complex I, III, and IV exhibit proton pump activity and, thus, contribute to the mitochondrial membrane potential (ΔΨ m ) across the IMM. The ΔΨ m represents the energetic force for the production of ATP by the F 0 F 1 ATPase that is also localized to the CM 7 . By knowing the spatial restriction of complexes I, III and IV to the CM, experimentally determined differences in pH of CM (pH 7.0) 8 and mathematically estimated pH values at the IBM (pH 7.4) 9 become comprehensible. The CJ, representing a physical diffusion barrier, constrains the diffusion of the protons to the cristae lumen. Rieger et al. showed that a pH gradient is even established between the Complex IV and the F 0 F 1 ATPase in the isolated CM compartment 8 . The proton motive force towards the matrix is the main component contributing to establish the ΔΨ m 10 . Wolf et al. showed recently that individual cristae are isolated insulators restricting the membrane potential generated at the CM by the CJ. This study demonstrated two distinct membrane potentials at the CM (ΔΨ CM ) and the IBM (ΔΨ IBM ) 11 .
Mitochondrial Ca 2+ uptake, which highly depends on ΔΨ, is essential to activate Ca 2+ -dependent dehydrogenases of the TCA cycle and increase oxidative phosphorylation 12 . Mitochondrial Ca 2+ Uniporter (MCU) complex-mediated Ca 2+ uptake is driven by the ΔΨ m across the IMM, yet it was never elaborated if the different CM (ΔΨ CM ) and the IBM (ΔΨ IBM ) potentials influence the Ca 2+ uptake on the level of mitochondrial ultrastructure.
The main regulator of MCU, mitochondrial Ca 2+ uptake 1 (MICU1), is localized at the IBM, stabilizes the CJ, and interacts with the MICOS complex 2,13 . Under basal conditions, MICU1 exists as a hexamer or oligomer, which in case of Ca 2+ binding via EF-hands of MICU1 disassembles into dimers with an EC 50 of 3.8 μM Ca 2+14, 15 . Methylation of MICU1 by protein arginine methyltransferase 1 (PRMT1) was shown to reduce the Ca 2+binding affinity of MICU1 to approximately 18.5 µM and inhibit thereby the Ca 2+ induced dissociation into dimers. Contrary, the specific binding of uncoupling protein 2 (UCP2) to methylated MICU1 counteracts the methylation effect and resensitizes MICU1 towards Ca 2+ (EC 50 = 4.0) 15 . Intracellular ER Ca 2+ release from the endoplasmic reticulum (ER) leads to OMM located Ca 2+ hotspots (up to 16.42 µM) in close proximity to mitochondrial-associated ER membranes (MAM), while mitochondrial regions without direct contact with the ER are facing much lower Ca 2+ concentration of approximately 2.94 µM 16 . The precise regulation of MICU1 by PRMT1 and UCP2 is important for cell metabolism and viability 17 . Especially cancer cells show a clear increase in UCP2 but also PRMT1 expression and both proteins when highly expressed are prognostic markers for lung carcinoma patients 18 .
The ability of MICU1 to respond to Ca 2+ hotspots with changes in its quaternary structure paired with the observation that MICU1 stabilizes CJ led us to hypothesize that the opening and stability of the CJ are directly regulated by high Ca 2+ 19 . Strong opening of the CJ is associated with a loss of membrane potential 2,11 , impaired oxidative phosphorylation and ATP production 20,21 , and induction of apoptosis 22 . Accordingly, we used structured illumination microscopy, differential membrane potential measurements of the CM and IBM, electron microscopy, and mitochondria-targeted genetically encoded biosensors for Ca 2+ to elaborate the involvement of MICU1 in CJ dynamics, effects on mitochondrial Ca 2+ uptake, and ΔΨ m homeostasis between IBM and CM.

Results
Spatial membrane potential distributions exist between CM and IBM. To detect spatial membrane potential gradients (SMPGs) between the CM (ΔΨ CM ) and the IBM (ΔΨ IBM ), we established a reliable and robust method using super-resolution microscopy. This approach builds on the assumption that the fluorescent potentiometric dye tetramethylrhodamine methyl ester (TMRM) is incorporated into the IMM 23-25 according to local ΔΨ, thus, the individual fluorescent intensities reflect distinct ΔΨ IBM and ΔΨ CM . While at low and moderate concentrations TMRM binds to the IMM in a linear correlation with the ΔΨ m , saturation effects have been reported at high TMRM concentrations 21 . Further, we could show that TMRM concentrations in the range of 1.35-81 nM show a slight saturation effect ( Supplementary Fig. 1). Therefore, we assumed that at a given high concentration TMRM gets saturated in the CM while the TMRM accumulation maintains in a linear correlation in the IBM with its lower ΔΨ. Thus, we hypothesized that both submitochondrial compartments vary in their labeling properties and the shifted TMRM affinity can be used to distinguish the different membrane potentials of the two sub-compartments of the IMM.
At very high concentrations of TMRM, self-quenching effects are known 25 . Therefore, we measured concentrations of 1. 35, 13.5, 81, 200, 500, and 1000 nM TMRM for self-quenching effects. For 1.35-81 nM no self-quenching effects were observed and only at 500-1000 nM TMRM we measured considerable selfquenching ( Supplementary Fig. 2), as reported elsewhere 26,27 . Accordingly, the used range of 1.35-81 nM TMRM used to detect mitochondrial membrane potential gradients was not affected by TMRM self-quenching effects.
To ensure that TMRM is specifically labeling and accumulating in the IMM, HeLa cells expressing mitochondrial matrix targeted superfolder green fluorescent protein (mt-sfGFP) and stained with 81 nM TMRM were treated with 100 µM ATP to induce mitochondrial swelling. A clear separation of TMRM cristae labeling and matrix mt-sfGFP fluorescence became obvious, showing specific labeling of the IMM by TMRM ( Supplementary  Fig. 3), as shown by others 23,24 .
In a first attempt, we investigated the sub-mitochondrial localization of TMRM and mt-sfGFP using super-resolution dualcolor structured illumination microscopy (SIM). At low TMRM concentrations (≤5.4 nM) the overlay of TMRM with mt-sfGFP shows an exclusive TMRM staining of the cristae membrane ( Fig. 1a). At TMRM concentrations ≥13.5 nM an additional halo of TMRM fluorescence surrounding the matrix develops ( Fig. 1a), pointing to stronger labeling of the inner boundary membrane (IBM).
Two different methods were used to quantify the spatial distribution of TMRM in relation to the reference channel mt-sfGFP: (1) The full width of half maxima (FWHM) of mitochondria for the TMRM and mt-sfGFP (i.e., matrix) channels were measured and their differences (ΔFWHM (TMRM − mt-sfGFP) ) illustrate the concentration-dependent TMRM expansion (Fig. 1b). (2) The IBM association index (AI IBM ) 2 , which correlates the TMRM fluorescence intensities of the central cristae membrane (CM) with the IBM (Fig. 1c). For both parameters, ΔFWHM (TMRM − mt-sfGFP) and AI IBM , a sigmoidal correlation between a TMRM accumulation in the mitochondrial periphery (halo) and increasing TMRM concentration was found, indicating different membrane potentials at the CM and IBM (Fig. 1d).
MICU1 and OPA1 contribute to the isolation of CM and IBM potentials. The CJ represents the bottleneck structure separating CM and IBM, thereby enabling high proton concentration gradients across the CM 11 . MICU1 2 and OPA1 4,28 are known to contribute directly to the stability of the CJ. Accordingly, the contribution of these proteins to the isolation of the different mitochondrial membrane potentials (ΔΨ m ) in CM (ΔΨ CM ) and IBM (ΔΨ IBM ) was investigated. Knockdown of either MICU1 29 ( Supplementary Fig. 4) Supplementary Fig. 5), thus, indicating that an intact CJ essentially needs either MICU1 or OPA1 and is fundamental to establish a high ΔΨ m restricted to the cristae. Investigations of the mitochondrial morphology revealed smaller and rounded mitochondria with lower interconnectivity in MICU1 and OPA1 depleted cells ( Supplementary Fig. 6). No mitochondrial swelling was observed upon depletion of either MICU1 or OPA1 and TMRM did not affect the mitochondrial morphology ( Supplementary Fig. 7).
Because UCP2 was recently shown to interact with methylated MICU1 and, thereby, pseudo-normalizing the Ca 2+ sensitivity and Ca 2+ dependent quaternary structure of methylated MICU1 15 , we next tested the roles of UCP2 and PRMT1 in CJ stability. As indicated by the low ΔFWHM (TMRM − mt-sfGFP) and AI IBM , knockdown of UCP2 18 15 it is tempting to speculate that UCP2 knockdown strengthens CM isolation from the IBM, since UCP2 does not interfere with methylated MICU1. This assumption is further supported by our findings that double knockdown of UCP2 and MICU1 prevented CM hyperpolarization upon UCP2 depletion ( MICU1, UCP2, and OPA1 knockdown influences spatial CM distribution. Recently, we reported that a knockdown of either MICU1 or OPA1 enlarge CJ width and affect cristae morphology 2 . To verify whether changes in IBM/CM TMRM ratio manifest due to the morphological changes in the IMM structure, the distribution of CM across the entire mitochondrial diameter was evaluated in MICU1, OPA1 or UCP2 depleted cells with transmission electron microscopy (Fig. 3a). MICU1 or OPA1 knockdown did not affect the ratio of the OMM to CM length (Fig. 3b). However, normalizing the cristae perimeter to the mitochondrial area, unveiled increased cristae density in OPA1 knockdown cells (Fig. 3c). OPA1 and MICU1 knockdown lead to mitochondrial fragmentation ( Supplementary Fig. 7) while the relative amount of CM in the mitochondria remains unchanged (Fig. 3b). The increased cristae density in OPA1 and MICU1 knockdown cells most likely relates to the decreased mitochondrial area/perimeter ratio. Silencing of UCP2 did not change cristae density.
However, the IBM/CM TMRM intensity ratios may also be influenced by the spatial distribution of CM within mitochondria. To evaluate this possibility, consecutive circular segments of mitochondria (shown in Fig. 3d) were analyzed in regard to the cristae density (ρ CM ). Both, MICU1 and OPA1 depletion result in a focused elevation of ρ CM in the mitochondrial center (Fig. 3e, f). Despite central clustering of IMM upon depletion of OPA1 or MICU1, both knockdowns lead to higher ΔFWHM (TMRMmt-sfGFP) and IBM association index (Fig. 2b, c). Because direct adaption of spatial membrane quantity would have led to increased CM fluorescence intensity under MICU1 or OPA1 knockdown, these results indicate that the spatial ρ CM distribution does not disturb the super-resolution measurements of ΔΨ IBM /ΔΨ CM . The knockdown of UCP2 causes a strong increase of ρ CM in the mitochondrial periphery (Fig. 3e, f), most likely derived from an increased number of CJ ( Supplementary Fig. 8).
Additionally, we evaluated whether or not the addition of TMRM in the various concentrations influences the cristae morphology or density. Neither the amount of CM nor ρ CM was changed while the spatial distribution of ρ CM was slightly elevated in the mitochondrial center with increased TMRM concentrations ( Supplementary Fig. 9).
Ca 2+ activation of MICU1 triggers a transient CJ opening and spatial cristae depolarization. MICU1 contains two EF-hands and undergoes structural rearrangements upon Ca 2+ binding 14 . Accordingly, we investigated the role of MICU1 in the opening of CJ upon Ca 2+ mobilization. Therefore, HeLa cells expressing an ER marker were stained with TMRM (5.4 nM), and the respective fluorescence intensities and distributions were recorded upon stimulation with histamine (Fig. 4a), a potent IP 3 -generating agonist. In this experimental setting, we were able to measure local variations of ΔΨ m depending on the proximity to the ER.
These experiments revealed that upon histamine stimulation HeLa cells react with local restricted losses of ΔΨ m in IMM regions with close proximity to MAMs and punctually along the entire IMM (Fig. 4b). The local TMRM spikes with a duration of several seconds indicate restricted fluctuations of the membrane potential of single cristae in close temporal correlation to cytosolic Ca 2+ elevations that trigger structural rearrangement of MICU1, destabilize the CJ and lead to instant local equilibration of ΔΨ CM and ΔΨ IBM .
We have recently shown that silencing of MICU1 destabilizes the CJ 2 and found herein that MICU1 knockdown leads to a Fig. 1 Measuring cristae and inner boundary membrane potential gradient. a Representative images of HeLa cells expressing mt-sfGFP (green), stained with 81, 40.5, 13.5, 5.4, 2.7, or 1.35 nM tetramethylrhodamine methyl ester (TMRM) (magenta) and examined using simultaneous dual-color 3D-SIM. b Intensity line plots of the mitochondrion in the white box in a showing the mt-sfGFP and TMRM intensity distribution. Full width at half maximum (FWHM) of mt-sfGFP and TMRM distributions are indicated with the corresponding maximal (I max ) and half-maximal (I max /2) intensities. c Schematic illustration of the calculation of the IBM association index on the example of HeLa cells expressing mt-sfGFP (green) and stained with 81 nM TMRM (magenta). The reference channel (mt-sfGFP) is thresholded and split by erosion and dilation into inner boundary membrane (IBM) and cristae membrane (CM) related segments, which are used as masks to measure TMRM mean intensities. The ratio of IBM intensity (I IBM,TMRM ) and CM intensity (I CM,TMRM ) results in the IBM association index. d Quantitative analysis of a using the ΔFWHM of TMRM and mt-sfGFP (described in b) or the IBM association index (described in c) to determine TMRM distribution to the IBM. The higher ΔFWHM or IBM association index, the broader the TMRM distribution indicating a stronger TMRM staining in the IBM. Data information: Horizontal lines in d represent the median, the lower and upper hinge show, respectively, first quartile and third quartile, and lower and upper whiskers encompass minimal and maximal values. Images and analyses were obtained from each 9-10 cells in 6 independent experimental days (n = 6). *P < 0.05 vs. 13.5 nM, # P < 0.05 vs. 40.5 nM and + P < 0.05 vs. 81 nM TMRM conditions evaluated using oneway analysis of variance (ANOVA) with Bonferroni post hoc test.
The role of UCP2 and Ca 2+ hotspots on CJ integrity/stability. We previously reported that UCP2 normalizes the reduced Ca 2+ sensitivity of PRMT1 methylated MICU1 and, thus, decreases its ability for Ca 2+ -induced fragmentation 14,15 . Knockdown of UCP2 in HeLa cells with high PRMT1 activity decreased the overall −Δ/+ΔTMRM response to histamine but maintained the amounts of spatial TMRM drops near MAMs (Fig. 4d). The reduced overall −Δ/+ΔTMRM ratio implies that the absence of UCP2 leads to a desensitization of MICU1 to Ca 2+ .

TMRM [nM]
Consequently, we conclude that in UCP2-depleted HeLa cells the CJ does not undergo conformational changes upon histamine stimulation, while the Ca 2+ hotspots in the MAM region are high enough to force MICU1 deoligomerization leading to an opening of the CJ and thereby homogenizes IBM and CM potential. Control experiments adding Ca 2+ buffer (CaB) instead of histamine showed no response in −Δ/+ΔTMRM and basal −Δ/+ΔTMRM ratio did not change by silencing either MICU1 or UCP2 compared to control cells (Fig. 4c, d).
Store-operated Ca 2+ entry (SOCE) after ER Ca 2+ depletion, which leads to a homogeneous cytosolic Ca 2+ elevation, was used to analyze −Δ/+ΔTMRM alterations ( Supplementary Fig. 10). After Ca 2+ re-addition no noticeable signals in −Δ/+ΔTMRM were detected for Control, MICU1, or UCP2 siRNA treated cells, supporting our assumption that only high local Ca 2+ concentration can lead to the opening of the CJ and spatial depolarization of the CM (Fig. 4d).
Spatial intensity of cytosolic Ca 2+ determines subcompartmental Ca 2+ propagation in mitochondria. The local depolarization of CM at MAMs upon stimulation with an IP 3generating agonist points to a CJ opening and permeability for ions into the cristae lumen (CL). Therefore, we tested the subcompartmental Ca 2+ kinetics in mitochondria upon IP 3 -induced ER Ca 2+ release and SOCE in MICU1, and UCP2 siRNA treated cells using red and green fluorescent genetically encoded Ca 2+ biosensors targeted either to the IMS and mitochondrial matrix, CL and mitochondrial matrix, or IMS and CL, respectively. The temporal delay of Ca 2+ propagation within the respective subcompartment of mitochondria was quantified as the delta of areas under the curve (ΔAUC) for normalized IP 3 -and SOCE-induced Ca 2+ signals in the respective sub-compartment (Fig. 5a-c).
By silencing MICU1, the delay between the Ca 2+ signal in the IMS and those in the CL and the matrix were reduced in SOCEbut not IP 3 -induced signals (Fig. 5d, e). The Ca 2+ propagation from the CL into the matrix was not affected during SOCE (Fig. 5f). Thus, the Ca 2+ propagation from the IMS to the cristae is under the control of MICU1 and the rate-limiting phase of the Ca 2+ propagation during SOCE. Nevertheless, the increased CL and decreased matrix Ca 2+ concentrations under SOCE in MICU1 knockdown cells indicate that MICU1 not only guards the CJ but also activates MCU-mediated Ca 2+ influx into the matrix ( Supplementary Fig. 11). Moreover, while the IP 3 -induced IMS to CL Ca 2+ -kinetics remained unaffected, the IMS to matrix and CL to matrix Ca 2+ kinetics were clearly accelerated by MICU1 silencing (Fig. 5d-f and Supplementary Fig. 12).
Silencing of UCP2 leads to a delayed propagation of Ca 2+ signals from the IMS to the CL and matrix after IP 3 -induced ER Ca 2+ release. In contrast, the Ca 2+ transition from the CL to the matrix was not affected (Fig. 5g-i and Supplementary Fig. 12). Furthermore, CL and matrix Ca 2+ delta intensities were reduced under the knockdown of UCP2 ( Supplementary Fig. 11), indicating a blockage of the CJ by UCP2 knockdown. For SOCE-induced mitochondrial Ca 2+ uptake, UCP2 silencing did not affect the Ca 2+ propagation between the various mitochondrial sub-compartments (Fig. 5g-i and Supplementary Fig. 12).
Taken together, these data indicate a biphasic Ca 2+ uptake mechanism that is dependent on the origin and spatial intensity of the cytosolic Ca 2+ signal and the stability of the CJ. In the first phase, IP 3 -induced ER Ca 2+ release leads to cytosolic Ca 2+ hotspot in the MAMs 16 . Subsequently, MICU1 deoligomerizes upon Ca 2+ binding at these hotspots leading to spatial CL depolarization due to a Ca 2+ induced opening of CJs (Fig. 4b) and consequently to the flooding of the CL with Ca 2+ , which, in turn, is instantly taken up into the matrix by constitutively active MCU within the CM. As Ca 2+ hotspots 16 and CJ openings (Fig. 4b) are short-lived, the second phase is dominated by the slower MICU1 mediated MCU shuttling into the IBM 2 achieving the second phase of Ca 2+ uptake into the mitochondrial matrix excluding the CL from the Ca 2+ uptake process (Fig. 5j). Notably, Cytosolic Ca 2+ signals originating from SOCE do not open the CJ due to a lack of Ca 2+ hotspots. Thus, the propagation of SOCE signals into the mitochondrial matrix relays only on the MCUshuttling due to partially deoligomerized MICU1 similar to the second phase in IP 3 -induced ER-Ca 2+ release (Fig. 5k).

Discussion
In this study, we combined dual-color SIM, TEM, and sophisticated live-cell imaging with mitochondrial sub-compartmental targeted genetically encoded biosensors to investigate the fundamental function of MICU1, OPA1, and UCP2 in the dynamic regulation of spatial membrane potential gradients (SMPGs) between IBM and CM. Moreover, the individual role of the two distinct IMM compartments (i.e. IBM and CM) in the orchestration of mitochondrial Ca 2+ uptake has been elucidated. The CM and the IBM are isolated by the CJ that consists of the "mitochondrial contact site and cristae organizing system" (MICOS), the optic atrophy 1 (OPA1) 28,32 and MICU1 2,13 .
Confirming recent results from ref. 11 , we herein demonstrate that the membrane potential of the IMM (ΔΨ IMM ) is separated by the CJ into two distinct potentials, that of the IBM (ΔΨ IBM ) and the CM (ΔΨ CM ). Notably, our data show that silencing of MICU1 leads to a strong shift of ΔΨ IBM /ΔΨ CM towards the IBM. This disruption of SMPGs points to a fundamental role of MICU1 in CJ integrity, thus, reemphasizing our previous results showing a CJ opening and clear reduction of the overall ΔΨ IMM for MICU1 knockdown conditions 2 . Moreover, our new data point to an alternative model of mitochondrial uncoupling that builds on the erosion of SMPGs by means of CJ-controlled ΔΨ CM . Opening of the CJ and the dissemination of the protons from the CL into the IMS yields depolarization of ΔΨ CM that mimics a mitochondrial uncoupling across the IMM.
We found in our recent study that the CJ widened through silencing of OPA1 or MICU1 2 . We assume that a destabilization Fig. 2 Cristae and inner boundary membrane potential gradient are controlled by MICU1, UCP2, and OPA1. a Representative images of HeLa cells expressing mt-sfGFP (green) and single or double transfected with Control, MICU1, OPA1, UCP2, or PRMT1 siRNA, stained with 81, 13.5, 5.4, or 1.35 nM tetramethylrhodamine methyl ester (TMRM) (magenta) and examined using simultaneous dual-color 3D-SIM. b Quantitative analysis of TMRM concentrations presented in a plus 40.5 and 2.7 nM TMRM concentration using the ΔFWHM of TMRM and mt-sfGFP to determine TMRM association with the IBM. The higher ΔFWHM, the broader the TMRM distribution indicating a stronger TMRM staining in the IBM. c Quantitative analysis of the same TMRM concentrations used in b using the IBM association factor of TMRM to determine TMRM association with the IBM. The relation of the IBM association factor to the TMRM distribution is similar to that of ΔFWHM b. Data information: horizontal lines represent the median, the lower and upper hinge show respectively first quartile and third quartile, and lower and upper whiskers encompass minimal and maximal values. Images and analyses were obtained from each 9-10 cells in 5-10 independent experimental days (n Control si = 8, n MICU1 si = 7, n OPA1 si = 5, n UCP2 si = 10, n MICU1 si /UCP2 si = 7, n PRMT1 si = 6, n PRMT1 si / UCP2 si = 6). *P < 0.05 vs. respective control si # P < 0.05 vs. respective MICU1 si and + P < 0.05 vs. respective UCP2 si conditions carried out with two-way analysis of variance (ANOVA) and Bonferroni post hoc test.
of the CJ leads to IMM constrictions which were shown to precede mitochondrial fission in ER close regions 19,33 . Our findings in this study suggest that mitochondrial fragmentation observed in MICU1 and OPA1 silenced cells is caused by destabilized CJ in ER close regions leading to IMM constriction and mitochondrial fission. Contrary reduction of UCP2 leads to a more stable CJ configuration 19 and thus to elongated mitochondria.
Because silencing of MICU1 destabilizes CJ under basal conditions and yields disruption of SMPGs, we speculate that even under resting condition, basal ER-originated Ca 2+ cycling within the MAMs achieves rare activation of MICU1 and, thus, increase in H + -permeability of the CJ. Besides the process of Ca 2+ cycling, CJ H + -permeability is further influenced by the methylation status of MICU1 by PRMT1. Methylation of MICU1 lowers the Ca 2+ binding affinity of MICU1 (EC 50 for Ca 2+ increases from 3.8 to 18.5 µM) 16 resulting in an enhanced stability/impermeability of the CJ. However, UCP2 binds exclusively to methylated MICU1, restores Ca 2+ sensitivity of methylated MICU1   (EC 50 = 4.0 µM) 15 , and pseudo-normalizes basal CJ H + -permeability. Consistently, in the present work, depletion of UCP2 resulted in an increased CJ stability due to the low Ca 2+ binding affinity of methylated MICU1 that cannot be activated by basal ER-originated Ca 2+ cycling within the MAMs in the absence of UCP2. In line with our findings, UCP2 knockdown is reported to increase ΔΨ m 34 , a phenomenon that, based on our present data, can be attributed to a stabilization of the CJ.  Under the condition of cell stimulation that triggers IP 3induced intracellular Ca 2+ release, high Ca 2+ at the mitochondrial surface binds to MICU1 yielding its deoligomerization, subsequent opening of the CJ, and disruption of individual SMPGs. Notably, the Ca 2+ concentrations achieved within the Ca 2+ hotspots in MAMs (up to 16.4 µM) during cell stimulation 16 , have the potential to disassemble methylated MICU1 (EC 50 for Ca 2+ 18.5 µM) 16 even in the absence of UCP2. Studies focusing on cristae dynamics revealed that CJ opening is predominantly restricted to MAMs after ER Ca 2+ release 30 and that CJs are closely associated with MAMs 35 . In line with this assumption, Gerencer and Vizi showed that initial mitochondrial Ca 2+ uptake after IP 3 -induced ER Ca 2+ release originates from local foci into the whole mitochondria 36 .
Recently we reported that a knockdown of MCU and EMRE leads to an enhanced accumulation of Ca 2+ inside the CL in response to IP 3 -induced Ca 2+ release 26 [40][41][42] . MICU1 was shown to potentiate MCU channel activity by direct interaction and MCU is not occluded by MICU1 38 , raising the question of how MICU1 reduces mitochondrial Ca 2+ uptake at low cytosolic Ca 2+ concentrations. These findings raise the question how MICU1 is regulating both, Ca 2+ availability in the CM and amplification of the conductance of MCU? Accordingly, based on the present findings, a 2-phase model of the orchestration of MCU-mediated mitochondrial Ca 2+ uptake that meets all reported features is proposed (Fig. 5j, k) Such sophisticated organization of the Ca 2+ -triggered opening of the CJ and biphasic mitochondrial Ca 2+ uptake ensures full mitochondrial functionality even in phases of high Ca 2+ challenges or to achieve uptake of low Ca 2+ . Due to local and temporal isolated CM depolarization, the mitochondria maintain general stability during Ca 2+ uptake while increasing the metabolic output. By the local regulation of CJ opening, not only Ca 2+ uptake is regulated but also the overall ΔΨ CM , mostly distanced from MAMs, is protected, and only affected in a minor way. This mechanism ensures mitochondrial stability under physiological Ca 2+ signaling.
Taken together, a model evolves by combining the separation of the mitochondrial membrane potential and the mitochondrial Ca 2+ uptake regulation by MICU1 at the structural level of the CJ (Figs. 4e and 5j, k). Both aspects are playing hand in hand, to finetune mitochondrial Ca 2+ and structural integrity to avoid matrix Ca 2+ deprivation or uncontrolled H + uncoupling through the CJ 19 . These findings have a profound influence on the understanding of mitochondrial metabolic regulation and how Ca 2+ signals are decoded into cellular responses.

Structured illumination microscopy (SIM)
Single and dual camera SIM imaging. The SIM setup used is composed of a 405, 488, 515, 532, and a 561 nm excitation laser introduced at the back focal plane inside the SIM box with a multimodal optical fiber. For super-resolution, a CFI SR Apochromat TIRF ×100-oil (NA 1.49) objective was mounted on a Nikon-Structured Illumination Microscopy (N-SIM®, Nikon, Austria) System with standard wide field and SIM filter sets and equipped with two Andor iXon3® EMCCD cameras mounted to a Two Camera Imaging Adapter (Nikon Austria, Vienna, Austria). At the bottom port, a third CCD camera (CoolSNAP HQ2, Photometrics, Tucson, USA) is mounted for wide-field imaging. For calibration and reconstruction of SIM images, the Nikon software (NIS-Elements, Nikon, Austria) was used. Reconstruction was permanently performed with the same robust setting to avoid artifact generation and ensures reproducibility with a small loss of resolution of 10% compared to the most sensitive and rigorous reconstruction settings. Microscopy setup adjustments were done as described elsewhere 2 .
Transfection procedures. HeLa cells were grown under standard culture conditions until 50% confluence was reached, transfected in DMEM (without FCS and antibiotics) with 1.5 µg/ml plasmids or 100 nM siRNA using 2.5 µg/ml TransFast™ transfection reagent (Promega, Madison, WI, USA). After 24 h, the medium was replaced with DMEM containing 10% FCS and 1% penicillin/streptomycin and kept for a further 24 h prior to experiments. The specific siRNAs (Microsynth, Balgach, Switzerland) used in this study are listed in Supplementary Table 1. Alternatively, plasmid transfection was done using PolyJet™ In Vitro DNA Transfection Reagent (SL100688, SignaGen® Laboratories, Frederick, MD, USA).
c is the molar concentration of TMRM, A the absorption of TMRM at 550 nm, ε is the specific extinction coefficient of TMRM at 550 nm in methanol, and l is the pathlength the light has to path through the TMRM solution.
Determination of TMRM quenching at various concentration. HeLa cells seeded on glass slides were labeled for 30 min with 1000, 500, 200, 81, 13.5, or 1.35 nM TMRM and transferred into a perfusion chamber. During the measurements cells were continuously perfused using a gravity-based perfusion system (NGFI, Graz, Austria) Measurements were performed on an inverted wide-field microscope (Observer.A1, Carl Zeiss GmbH, Vienna, Austria) as described previously 15 . TMRM was excited at 550 nm, and the emission was collected at 600 nm. A full disruption of the mitochondrial membrane potential was done by the application of carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) (Abcam, Cambridge, UK). In the case of the non-quenching mode of TMRM, a direct drop of mitochondrial fluorescence followed FCCP treatment. In the case of quenching mode, a clear increase in TMRM fluorescence after FCCP addition was observed, followed by a drop in intensity. During the entire experiment the respective TMRM concentrations were present within the imaging buffers and prior recording the cells were equilibrated in the perfusion chamber for a further 5 min. Mitochondrial TMRM signals of single cells were background corrected using a background ROI.
Determination of TMRM saturation at various concentration. HeLa cells seeded on glass slides were labeled for 30 min with 81, 40.5, 13.5, 5.4, 2.7, or 1.35 nM TMRM and transferred into a perfusion chamber containing the same concentration of TMRM. Per coverslip, 40 cells were imaged randomly and the average cellular TMRM fluorescence was measured.
Cristae membrane and inner boundary membrane potential separation by FWHM.
HeLa cells transfected with mt-sfGFP were stained with different concentration of TMRM ranging from 1.35 to 81 nM. After the staining procedure the cells were kept in loading buffer containing the respective concentrations of TMRM and imaged with dual-color 3D-SIM. To compensate for intensity differences for TMRM acquisition the laser power was adjusted to match image intensity histograms of different TMRM concentrations. Post imaging, the recorded data were background corrected using an ImageJ Plugin (Mosaic Suite, background subtractor, NIH) with a sliding rectangle diameter of 50 pixel. Intensity line plots of mitochondrial mt-sfGPF and TMRM fluorescence were manually measured with a width of 50 pixels (1.6 µm). The FWHM of mt-sfGFP and TMRM fluorescence distributions was measured using linear interpolation to gain subpixel information. Subtraction of FWHM mt-sfGFP from FWHM TMRM results in ΔFWHM.
Cristae membrane and inner boundary membrane potential separation by IBM association index. The IBM association factor of TMRM or MTG was calculated as described elsewhere 2 . In short, images were subjected to background subtraction (Mosaic Suite, background substractor, NIH) with a sliding rectangle diameter of 50 pixels. The reference channel (sf-GFP, mtDsRed, or MTG) was Otsu 45 auto thresholded and further dilated and eroded in two independent subsets. One erosion and two dilation iterations were used. Pixel-wise subtraction of the erosion reference of the dilated reference image yields a hollow structure, used as a mask to measure the mean intensity in the mitochondrial periphery or IBM-related area in the object channel. The erosion reference served as a mask to measure the bulk or cristae mean fluorescence intensity. The ratio of IBM/CM mean intensity is a value to estimate changes in the object label distribution inside a mitochondrion, which is referred to as the IBM association index. The higher the ratio value the higher the distribution of protein labels in the IBM. For image analysis the freeware program ImageJ was used.
Morphological analysis of mitochondria in 2D-SIM images. 3D-SIM and timelapsed images of EMRE-mCherry, MTR, MTG, or MICU1-YFP were used for morphological analysis. Images were background corrected with an ImageJ Plugin (Mosaic Suite, background subtractor, NIH) and the subsequent binarization was done using an Otsu 45 auto threshold. The ImageJ particle analyzer was used to extract the mitochondrial count (c), area (a), perimeter (p), minor (x), and major (y) axes of the mitochondria. Aspect ratio (AR) was determined as The Form Factor (FF) was determined as followed: Electron microscopy. Cells were incubated in a loading buffer with 81, 13.5, or 1.35 nM TMRM for 30 min, washed with PBS, fixed with 2.5% glutardialdehyde and 2% formaldehyde in a buffered solution, and postfixed in either 2% osmium tetroxide or 1% osmium tetroxide that had been reduced with 1% potassium hexacyanoferrate 46 . The cells were dehydrated in an ascending ethanol series, embedded in TAAB embedding resin, and sectioned on a Leica Ultracut 7 ultramicrotome using a Diatome diamond knife. The sections were counter-stained using platinum blue (IBIlabs, Boca Raton, FL, USA) and lead citrate (Leica Microsystems, Wetzlar, Germany) and visualized in an FEI Tecnai 20 transmission electron microscope. They were photographed at ×2000 magnification with a Gatan ultrascan 1000 camera.
Analysis of cristae membrane and density. In a first step, mitochondria were segmented inside the micrographs manually by free hand selection. Afterwards, the segmented mitochondria were isolated and the cristae inside each mitochondrion was segmented by free hand selection in ImageJ. The process was semi-automated using ImageJ macros. Mitochondrial and cristae area and perimeter were measured. The perimeter of mitochondria divided by the cristae perimeter gives an indication of membrane alteration of the cristae. The ratio of cristae perimeter divided by mitochondrial area gives a representation of cristae density.
Analysis of cristae density distribution. For analysis of the spatial cristae density distribution, segmented mitochondria and respective cristae were selected.
Segmented mitochondria were binarized and used as a mask for binarized cristae membranes to determine the percentage coverage of cristae perimeter inside the mitochondria area. Iteratively, the mitochondrial mask was eroded homogeneously in small increments of 2-pixel width (5.88 nm) and the respective coverage of cristae was measured. The result of these measurements represented a cristae density in circular segments starting from the outer mitochondrial membrane towards the mitochondrial center. As mitochondrial size and shape vary that results in different numbers of ring segments, the cristae densities were normalized by linear interpolation to 100 segments.
Analysis of local drops in membrane potential in close proximity to MAMs. Cells were transfected with an insensitive ER targeted ATP probe ERAT4.03 NA (NGFI, Graz, Austria 47,48 ) as a marker for the ER and stained with 13.5 nM TMRM in loading buffer for 30 min. TMRM was present during the entire experiment. Cells were afterwards transferred to the live cell chamber containing 0.8 ml loading buffer and imaged over time with a frequency of 0.25 Hz. 8 s after image sequence initialization 0.8 ml of loading buffer with or without 200 µM histamine was added to a final concentration of 100 µM histamine.
To analyze whether local drops of membrane potential appear under SOCE, cells were incubated with 2 µM Fura-2AM (MoBiTec GmbH, Göttingen, Germany) and 13.5 nM TMRM in loading buffer for 30 min. TMRM was present during the entire experiment. Cells were challenged in Ca 2+ -free buffer with 100 µM histamine/10 µM CPA (Sigma-Aldrich, Vienna, Austria) to empty ER-Ca 2+ stores and imaged using the widefield setup with 380 nm illumination and widefield acquisition with the bottom port CCD camera. Afterward, the same cells were imaged over time with a frequency of 0.25 Hz using the dual-cam SIM setup. 8 s after image sequence initialization Ca 2+ was added to a final concentration of 2 mM.
A custom-made ImageJ macro was used to analyze spatial drops in mitochondrial membrane potential in dependency of spatial proximity to ERrelated mitochondrial-associated membranes (MAM) and ER Ca 2+ release. The analysis was separated into several steps, (1) definition of MAMs, (2) identification of drops in the mitochondrial membrane potential, and (3) spatial assignment of the drops to MAM and non-MAM mitochondrial areas.
(1) After acquisition and reconstruction using NIS-Elements AR the mitochondrial TMRM channel (further referred to as C M ) and the ERAT4.03 NA channel (further referred to as C E ) were background subtracted (Mosaic Suite, background subtractor, NIH) with a sliding rectangle diameter of 50 pixels. Both channels (C M and C E ) were bleaching corrected using the histogram matching function, which is implemented in Fiji. An Otsu auto threshold was used to binarize C M and C E . The overlap of both channels was eroded (iteration = 1; count = 1) to remove small structures and subsequently dilated (iteration = 6; count = 1) and used as a mask for C M to assign areas of mitochondria with close proximity to MAMs. (2) Background subtracted (Mosaic Suite, background subtractor, 50 pixels, NIH) and bleaching corrected (histogram matching) C M was analyzed for frame-to-frame intensity changes (ΔC M ). Additionally, Otsu auto thresholded C M of both images involved for the calculation of ΔC M was used as a mask for ΔC M to remove the mitochondrial movement from ΔC M . Negative changes from one frame to the next (−ΔC M ) as well as positive ones (+ΔC M ) were subjected to a threshold of 35% of the original intensity of the first frame. Thresholded −ΔC M and +ΔC M were eroded (iteration = 1; count = 1) and subsequently dilated (iteration = 1; count = 1) to remove remaining halos around mitochondria still originating form mitochondrial movement. Finally, the area of −ΔC M and +ΔC M containing segments were quantified and the ratio of −ΔC M /+ΔC M was calculated. For more easy interpretation, −ΔC M and +ΔC M were replaced with −ΔTMRM and +ΔTMRM, respectively. The area of −ΔC M quantifies distinct local losses of TMRM signal, which can be induced by direct loss of membrane potential or by the movement of the cristae membrane itself. Shifts in the lateral or axial direction within mitochondria can lead to a loss of TMRM fluorescence. Nevertheless, cristae membrane movement should account for increases in TMRM signals in the same quantity. A ratio of −Δ/+ΔTMRM = 1 states that equal amounts of distinct changes, increase or loss of TMRM signal, are present and can be referred to a steady state in which movement of cristae membrane is the main contribution to −ΔTMRM and +ΔTMRM. −Δ/+ΔTMRM < 1 shows more increases in TMRM signal indicating the generation of mitochondrial membrane potential. −Δ /+ΔTMRM > 1 on the other side implies a local high loss of membrane potential. The whole mitochondrial area derived from auto Otsu binarized CM as well as MAMs determined in (1) were used as a mask for −ΔTMRM and +ΔTMRM areas to differentiate −Δ/+ΔTMRM-ratios in regard to the proximity to MAMs.
Technical note. Lateral and axial movements of the mitochondria and the CM over time may cause errors in the readings of the intensiometric TMRM signals. To counter this issue, the delta TMRM fluorescence (ΔC M ) of subsequent frames was calculated and thresholded by ± 35% of the fluorescence intensity change per frame and per pixel and classified them into positive (+ΔTMRM) and negative (−ΔTMRM) area partitions. Accordingly, the ratio of −Δ/+ΔTMRM compensates for errors due to CM movements and quantifies local variations of membrane potential of the IMM. The ER-marker was used to identify mitochondrial structures in close proximity to the ER (MAMs) (Fig. 4a). The ratio of −Δ/+ΔTMRM was calculated over time.
Ca 2+ imaging experiments. Ca 2+ imaging was performed as described elsewhere 49 . In brief, Ca 2+ was measured on a digital wide-field microscope, the iMIC (Till Photonics, Gräfelfing, Germany) equipped with a 40-objective (alpha Plan Fluar 40, Zeiss, Göttingen, Germany) and an ultrafast switching monochromator, the Polychrome V (Till Photonics). Illumination of GEM-GeCO1 targeted sensors was performed at 430 nm excitation and emissions were collected with a dichrotome dual emission filter set (dichroic 535dcxr). For dual recordings, GEM-and CAR-GeCO1 targeted sensors were alternately excited for 500 ms each at 430 and 575 nm. Emissions derived from both sensors were taken in 3 s intervals. During the measurements, cells were continuously perfused using a gravity-based perfusion system (NGFI, Graz, Austria) and images were recorded with a chargecoupled device (CCD) camera (AVT Stingray F-145B, Allied Vision Technologies, Stadtroda, Germany). Data acquisition and control of the digital fluorescence microscope were performed using the live acquisition software version 2.0.0.12 (Till Photonics). For analysis, fluorescent traces of single cells were background corrected using a background ROI, and bleaching was corrected using an exponential decay function in a costume made excel macro. To compare the kinetics of two mitochondrial compartments, the fluorescent traces were normalized to the minimum and maximum of the IP 3 or SOCE-induced mitochondrial Ca 2+ elevations. The area under the curve was determined for both compartments using the trapezium method. The delta of both AUCs was calculated. The starting point of the ΔAUC was set by the point at which the Ca 2+ signal in one of both compartments surpasses the three-fold of the standard deviation of the basal signal. The end of the ΔAUC was set to the time point at which one of the signals reached the maximum and crosses with the intensity of the opposite Ca 2+ trace from the other sub-mitochondrial compartment.
Statistics and reproducibility. Each exact n value and the number of independent experiments are indicated in each figure legend. Statistical analysis was performed using the GraphPad Prism software version 5.04 (GraphPad Software, San Diego, CA, USA) or Microsoft Excel (Microsoft Office 2013). Analysis of variance (ANOVA) with Bonferroni post hoc test and t test were used for evaluation of the statistical significance. All box plots show minimum to maximum values if not otherwise indicated. The central line is the median with boxes extending to 25 and 75% and the whiskers encompass all data. P < 0.05 was defined to be significant. At least three experiments on 3 different days were performed for each experimental set-up.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the authors on reasonable request, see author contributions for specific data sets.

Code availability
The code used in this study is available from the authors on reasonable request; see "Author contributions" for specific code sets.